A method and system for high-throughput measurement of dynamic growth phenotypes of plant root systems

Abstract

The application provides a kind of high-throughput plant root system dynamic growth phenotype measurement method and system, belong to plant phenomics and computer vision cross technical field, including obtaining plant root system time series image;Select calibration image, generate calibration instruction set;Reference calibration instruction set image corrects image data in high-throughput time series image data set. Utilize time series difference method to process corrected sub-image sequence, pixel-level accumulation is carried out to the image sequence after difference, and the static accumulation image in observation period is obtained;The static accumulation image is adaptively binarized and segmented, and the single-pixel-width centerline skeleton is extracted, to obtain the skeleton image;The skeleton image is processed using the shortest path routing algorithm, and the main root length is obtained. Based on the quantified main root length at each time point, a growth curve graph of the main root length over time is generated, which realizes the quantitative analysis of the root growth dynamics and provides dynamic plant root change information for subsequent research.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of plant phenomics and computer vision, specifically relating to a high-throughput method and system for measuring the dynamic growth phenotype of plant roots. Background Technology

[0002] In modern life science research and agricultural breeding practices, high-throughput phenotyping has become a core driving force. Through automated and large-scale methods, it enables the rapid and accurate acquisition of morphological, physiological, and biochemical characteristics of large numbers of biological samples to meet the urgent needs of screening massive genetic resources, analyzing gene function, and studying environmental interactions. Plant roots, as key organs for life activities, have root system architecture (RSA) and its dynamic changes that are crucial phenotypes determining plant productivity. Therefore, high-throughput phenotyping of roots is currently a technological hotspot and challenge. Accurate, non-destructive, and high-throughput quantification of dynamic parameters such as taproot growth rate is an important aspect of crop breeding, plant physiology, and agricultural research.

[0003] Traditional measurement methods include manual measurement and destructive sampling, semi-automated analysis based on general-purpose image software, and precision imaging analysis systems based on high-end equipment. Traditional manual measurement and destructive sampling are highly destructive, directly causing plant death during excavation and making dynamic observation impossible. They are extremely labor-intensive and inefficient, unsuitable for large-scale sample screening. Furthermore, they suffer from poor measurement accuracy and repeatability; manual measurement easily introduces subjective errors, and the breakage and loss of small roots during excavation is almost unavoidable, leading to severely distorted measurement results. Semi-automated analysis based on general-purpose image software has low automation and is highly dependent on human intervention. For time-series data containing hundreds of images, manual or semi-manual parameter adjustments and plotting operations are performed image by image. General-purpose image software cannot correct geometric distortions. When using a camera for overhead shooting, it is difficult to ensure the lens is perfectly perpendicular to the petri dish plane; the resulting perspective distortion will systematically underestimate the measured root length of non-central roots, affecting the accuracy of the measurement data. Precision imaging analysis systems based on high-end equipment, such as CT and MRI equipment, are expensive and require specialized operating environments and personnel, making them difficult to implement in ordinary laboratories. Meanwhile, its data processing flow is complex and time-consuming, and the three-dimensional reconstruction and analysis process has extremely high requirements for computing resources. For research scenarios that only need to focus on specific two-dimensional dynamic parameters such as the elongation of the main root, it actually reduces the processing throughput.

[0004] Whether relying on direct measurement using manual scales or semi-automated analysis using general-purpose image processing software, the "single sample, single image" processing mode is inadequate when dealing with high-throughput data, failing to meet the demands of modern scientific research for data processing efficiency and scale. Therefore, a measurement method is needed that can process data on the dynamic growth process of primary roots in batches while ensuring accurate and consistent results in large-scale data analysis. Summary of the Invention

[0005] To address the shortcomings of existing technologies in the quantitative analysis of plant root dynamics, this invention provides a high-throughput method and system for measuring dynamic growth phenotypes of plant roots.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A high-throughput method for measuring the dynamic growth phenotype of plant roots includes the following steps: Acquire a high-throughput time-series image dataset of plant roots; select calibration images and generate a calibration instruction set; the calibration instruction set includes a custom identifier for each petri dish in a single image, the coordinates of the four vertices of the reference object, the global coordinates of the analysis region, and the physical size information of the reference object; Based on the parameters of the reference objects in the calibration instruction set, the image data in the high-throughput time-series image dataset is processed to obtain the corrected sub-image sequence; The corrected sub-image sequence is processed using the temporal difference method to obtain a difference image sequence; the difference image sequence is accumulated pixel-level to obtain a static accumulated image within the observation period; adaptive binarization segmentation is performed on the static accumulated image, and a centerline skeleton with a single pixel width is extracted based on the image after adaptive binarization segmentation to obtain a skeleton image; the shortest path algorithm is used to traverse all skeleton segments in the skeleton image to obtain the principal root length obtained by quantization at each time point; Based on the quantified taproot length at each time point, a growth curve of taproot length over time is generated, enabling quantitative analysis of root growth dynamics.

[0007] Preferably, the step of selecting a calibration image and generating a calibration instruction set specifically includes the following steps: Obtain a representative calibration image containing one or more culture containers; through the interactive interface, mark a reference object with at least four identifiable coplanar vertices on the image for each culture container, input the actual physical size of the reference object, and record the pixel coordinates of the four vertices of the reference object in the original image; Based on the pixel coordinates of the four vertices of the reference object and the coordinates of a preset standard target rectangle, a transformation matrix for correcting perspective distortion is obtained. Based on the transformation matrix, an analysis region for root growth is accurately selected. A matrix for inverse transformation is calculated, and the selected analysis region for root growth is converted back to its corresponding position in the original image coordinate system using the inverse transformation matrix. Save the user-defined identifier for each petri dish, the coordinates of the four vertices of the reference object, the coordinates of the analysis area, and the physical size information of the reference object as a calibration instruction set.

[0008] Preferably, the step of using the shortest path algorithm to traverse all skeleton segments in the skeleton image and obtain the quantized principal root length at each time point specifically includes the following steps: Connectivity analysis was performed on the skeleton to identify all independent skeleton segments; Calculate the farthest Euclidean distance between any two points in the skeleton image, and use the two points obtained as candidate start and end points of the path; The shortest path algorithm is used to calculate the shortest distance between the starting point and the ending point under the constraints of the skeleton segment; all skeleton segments are traversed, and the longest path calculated is identified as the main root path of the plant. The number of pixels contained in the main root path is counted, and the length of the main root is calculated.

[0009] Preferably, the static accumulated image is subjected to adaptive binarization segmentation using the maximum inter-class variance method.

[0010] Preferably, the image data in the high-throughput time-series image dataset is processed based on the parameters of the reference objects in the calibration instruction set to obtain a corrected sub-image sequence. Specifically, the samples are batch-perspective corrected and segmented using a double-loop structure. The outer loop reads each frame of the high-throughput time-series image containing multiple culture dishes, and the inner loop reads the parameters of the corresponding calibration instruction set to obtain different plant sample IDs, reference vertex source coordinates, and global coordinates of the analysis region. For each sample ID, a forward perspective transformation matrix is ​​calculated based on the reference vertex source coordinates and the coordinates of a preset target rectangle. The high-throughput time-series image is then processed using the forward perspective transformation matrix to crop out irregular areas of the sample and correct them into standard rectangular sub-images. The global coordinates of the analysis region of the sample are mapped from the original image coordinate system to the standard coordinate system coordinates of the corrected sub-image using the perspective transformation matrix to obtain the corrected sub-image sequence.

[0011] Preferably, before generating the growth curve of the main root length changing over time, the method further includes calculating a scale factor based on the parameters of the reference object of the calibration instruction set and the correction target pixel size, converting the pixel length of the skeleton segment identified as the main root into the physical length of the main root, and generating the growth curve of the main root length changing over time based on the physical length of the main root.

[0012] Preferably, the method further includes performing a morphological closing operation on the binary image before extracting the centerline skeleton of a single pixel width based on the binary image to obtain an enhanced binary image, and extracting the centerline skeleton of a single pixel width from the enhanced binary image; wherein, performing the morphological closing operation on the binary image specifically involves performing three consecutive dilation operations on the binary image, and then performing three consecutive erosion operations on the dilated binary image to obtain the enhanced binary image.

[0013] This invention also provides a high-throughput measurement system for dynamic plant root growth phenotypes, specifically comprising: The image calibration module acquires a high-throughput time-series image dataset of plant roots; selects a calibration image and generates a calibration instruction set; the calibration instruction set includes a custom identifier for each petri dish in a single image, the coordinates of the four vertices of the reference object, the global coordinates of the analysis area, and the physical size information of the reference object.

[0014] The correction segmentation module is used to process the image data in the high-throughput time-series image dataset based on the parameters of the reference objects in the calibration instruction set, so as to obtain the corrected sub-image sequence.

[0015] The trajectory extraction module is used to process the corrected sub-image sequence using the temporal difference method to obtain a difference image sequence; to perform pixel-level accumulation on the difference image sequence to obtain a static accumulated image within the observation period; to perform adaptive binarization segmentation on the static accumulated image, and to extract a centerline skeleton with a single pixel width based on the image after adaptive binarization segmentation to obtain a skeleton image; and to traverse all skeleton segments in the skeleton image using a shortest path traversal algorithm to obtain the principal root length obtained by quantization at each time point.

[0016] The length quantification module generates a growth curve of the main root length over time based on the quantified main root length at each time point, thereby realizing a quantitative analysis of the root growth dynamics.

[0017] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps described in the method for measuring the dynamic growth phenotype of plant roots.

[0018] The present invention also provides a computer-readable storage medium storing a computer program that, when loaded by a processor, can execute the steps described in the method for measuring the dynamic growth phenotype of plant roots.

[0019] The high-throughput method for measuring the dynamic growth phenotype of plant roots provided by this invention has the following beneficial effects: The processing method proposed in this invention, after selecting calibration images and generating a calibration instruction set, allows for fully automated processing of hundreds or thousands of time-series images without manual intervention, significantly improving processing efficiency and enabling highly efficient high-throughput root phenotypic analysis. By comparing the image data in the high-throughput time-series image dataset with reference data in the calibration instruction set, corrected sub-image sequences are obtained. Systematic geometric errors caused by inconsistent camera shooting angles are eliminated, ensuring that all samples are measured on the same scale. Through time-series differencing and trajectory accumulation strategies, the analysis object is transformed from the static root system as a whole to the dynamic growth trajectory, enabling more precise capture and quantification of the root growth process. This analysis of "process" rather than "state" is not only conceptually closer to the essence of biological growth but also technically more robust due to its effective filtering of static background noise. Adaptive binarization segmentation of the static accumulated images automatically addresses differences in brightness and contrast between different samples and images at different time points, avoiding the subjectivity and inconsistency caused by manually setting thresholds. Based on the quantified taproot length at each time point, a growth curve of taproot length over time is generated, enabling quantitative analysis of root growth dynamics and obtaining a complete growth trajectory map. This achieves effective extraction and reconstruction of the growth process, and enables quantitative monitoring of root dynamic growth processes in time-series image sequences. Attached Figure Description

[0020] To more clearly illustrate the embodiments and design schemes of the present invention, the accompanying drawings required for this embodiment will be briefly described below. The drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the hardware device according to an embodiment of the present invention.

[0022] Figure 2 The diagram shows the algorithm processing results in an embodiment of the present invention.

[0023] Figure 3 This is a schematic diagram illustrating the change in the length of the primary root in an embodiment of the present invention.

[0024] Figure 4 This is a flowchart of a high-throughput method for measuring the dynamic growth phenotype of plant roots according to the present invention. Detailed Implementation

[0025] To enable those skilled in the art to better understand and implement the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0026] The time-series images required for automated analysis in this invention can be acquired using a high-throughput automated imaging device. For example... Figure 1 As shown, the overall structure of the device is as follows: In a temperature- and humidity-controlled darkroom environment, a large number of transparent petri dishes containing plant seedlings are vertically cultured in a standardized array using a high-density culture module. A high-resolution camera deployed in front of the culture module has its shooting cycle precisely synchronized with the activation cycle of a time-controlled lighting system. In this way, the system can automatically acquire high-quality images of all samples at fixed time intervals (e.g., once per hour) during brief periods of illumination, and then restore the dark environment, thereby obtaining a high-throughput time-series image dataset in a lossless and automated manner that can be directly batch-processed by the subsequent methods of this invention.

[0027] This invention provides a method for measuring the dynamic growth phenotype of plant roots using a dataset acquired by a high-throughput automated imaging device. Figure 4 As shown, the specific steps include: Step 1: Generate a batch processing instruction set for image calibration, which includes the following steps: S11. Calibration Parameter Acquisition: Obtain the global coordinate set of the reference object vertices in the original image. For each culture container, through the human-computer interaction interface, the operator marks a reference object (e.g., a petri dish or a square label) with at least four identifiable coplanar vertices on the image and inputs the actual physical dimensions of the reference object. The pixel coordinates of the four vertices of the reference object in the original image are recorded through a mouse callback function. The two-dimensional pixel coordinates of these four points are then... , ),( , ),( , ),( , The first set of coordinates is formed by combining the coordinates of the first set of coordinates, denoted as . This set represents the actual position of the reference object in the original image coordinate system.

[0028] S12. Define the standard vertex coordinate set of the target correction plane. A virtual target correction plane is predefined; this plane has a standard geometric shape, typically with a predefined width. and height A standard rectangle can be pre-defined according to application requirements. The coordinates of the four vertices of this standard rectangle in a two-dimensional Cartesian coordinate system are determined to form the target coordinate point set, denoted as [the set of points]. The set D = { To ensure the correctness of the mapping, the set The order of the vertices in the middle must match the set. The order of the vertices in the middle corresponds one-to-one. For example, ( , ) is the top left vertex of the source quadrilateral, ( , () represents the top-left vertex of the target rectangle. This set represents the position of the reference object in an ideal, distortion-free "bird's-eye view".

[0029] S13. Virtual Correction and Precise Positioning of the Analysis Area: To accurately select the analysis area in a distortion-free view, a forward transformation matrix for correcting perspective distortion is calculated based on the pixel coordinates of the four vertices of the reference object and the coordinates of a preset standard target rectangle. A temporary, distortion-free "bird's-eye view" is then generated based on this matrix. The root growth analysis area is then precisely selected from the bird's-eye view. The specific calculation steps for the forward perspective transformation matrix are as follows:

[0030] (1) Set the source coordinate points and target coordinate point set The input parameters are provided to a computational unit. This unit, based on the principle of homography, solves for a 3x3 perspective transformation matrix. The matrix The source coordinate point set is uniquely defined. Accurate mapping of any point in the target coordinate set The geometric transformation relationship of corresponding points in the middle.

[0031] This transformation relationship follows the mathematical model of homogeneous coordinates: ; in: These are the coordinates of any point in the source image. It is its corresponding coordinate in the target plane. It is a non-zero scale factor. This is the 3x3 perspective transformation matrix to be solved. The computational unit utilizes... and By identifying four pairs of corresponding points, a system of linear equations is established and solved to calculate the matrix. Eight independent parameters.

[0032] (2) Apply the transformation matrix to generate the corrected image, using the original image (or the image region containing the source quadrilateral) as the input image, and the calculated forward perspective transformation matrix. As the basis for transformation, and using the dimensions of the defined standard rectangle... The dimensions are used as the output image size. A perspective transformation operation (Warp Perspective) is performed on the input image, i.e., based on the matrix... The defined mapping rules recalculate and rearrange each pixel within the source quadrilateral region, filling it into the corresponding position in the output image. This operation ultimately generates a temporary image of size [size missing]. A corrected image without perspective distortion, also known as a "bird's-eye view".

[0033] S14. Inverse Coordinate Transformation: Calculate an inverse perspective transformation matrix for the inverse transformation, and select the analysis region of root growth. Then, use the inverse transformation matrix to convert the selected region back to its corresponding position in the original image coordinate system. The specific calculation steps for the inverse perspective transformation matrix are as follows:

[0034] (1) Call a computing unit to obtain the standard vertex coordinate set of the target correction plane. As the source point set, and the global coordinate set of the vertices of the original image reference. As the destination points, calculate a 3x3 homography matrix. This matrix is ​​the desired inverse perspective transformation matrix, denoted as . This matrix uniquely defines the perspective mapping from the target correction plane to the original image plane.

[0035] (2) For any point on the target correction plane The point corresponding to it on the original image The following homogeneous coordinate transformation formula can be used for calculation: ; in, These are the coordinates of a point on the target correction plane (i.e., the "bird's-eye view"). These are the coordinates of the corresponding point in the original image coordinate system, obtained through transformation. It is a non-zero homogeneous coordinate scaling factor. This is the 3x3 inverse perspective transformation matrix to be solved, and its elements are... to These are the parameters of the matrix. Typically, they are... Normalized to 1, the matrix has 8 unknown degrees of freedom.

[0036] Expanding the matrix multiplication above, we can obtain the formula for calculating Cartesian coordinates: ; ; (3) Calculation unit utilizes and Given four pairs of points (a total of 8 independent equations), establish a system of linear equations and solve for the matrix. 8 unknown parameters This uniquely determines the transformation matrix.

[0037] S15. Calibration instruction set file generation: The user-defined identifier for each petri dish, the coordinates of the four vertices of the reference object, the global coordinates of the analysis area (results from S14), and the physical size information of the reference object are all saved together as a structured global calibration file, namely the "calibration instruction set" (JSON file). The generation of this file signifies that all steps requiring manual intervention have been completed, and subsequent steps will enter a fully automated batch processing stage.

[0038] Step Two: Batch Correction and Segmentation of Image Sequences. A high-throughput time-series image dataset of plant roots is acquired and used as input. The calibration instruction set generated in Step One is used to process the image data in the high-throughput time-series image dataset, obtaining corrected sub-image sequences to eliminate perspective distortion and establish a unified measurement scale. The specific implementation is as follows:

[0039] S21. Scale factor calculation: Based on the actual physical size of the reference object recorded in the calibration file, and the corrected target pixel size set for the reference object in step one. Calculate a scaling factor that represents the correspondence between physical size and pixel size. = The calculated scale factor Stored as a global calibration parameter. This scaling factor will be used in the subsequent automated batch processing of all samples. It is used as a consistent scale benchmark, serving as the basis for converting all subsequent pixel measurements into true physical length. For pixel measurements obtained on any corrected sub-image... (For example, root length) can be determined by dividing by this scaling factor. To accurately convert it back to its actual physical length .

[0040] S22, Batch Spatial Segmentation and Perspective Correction, this process is implemented through a double loop structure. First, an outer loop is started to traverse each frame of the original image in the high-throughput time-series image dataset (the original image contains multiple petri dish samples at the same time). For each frame of the loaded original image, an inner loop is started to traverse each independent petri dish sample entry (e.g., "ID 1", "ID 2", ...) recorded in the "calibration instruction set" (JSON file). In the inner loop, for a specific sample ID, the following is performed: (1) Coordinate extraction, retrieving and extracting the coordinates of the four reference vertex points (i.e., the source coordinate point set) corresponding to the specific sample ID from the calibration instruction set. (2) Matrix calculation, using the coordinates of the four vertex points of the reference object in the current image. and the preset target rectangle coordinates Calculate a forward perspective transformation matrix that applies only to this single sample. The specific calculation method is the same as step S13(1). (3) Transformation and extraction: Call the perspective transformation function to transform this matrix. This operation is applied to the entire original image. Functionally, it achieves spatial segmentation, i.e., based on the matrix. The definition is to precisely "crop" the irregular quadrilateral area occupied by the specific sample from the original large image, and at the same time "stretch" and correct it into an independent, standard, perspective-distortion-free rectangular sub-image with a pixel size in a fixed ratio to the physical size.

[0041] S23. Normalization and Sub-image Sequence Reconstruction: This double loop iterates continuously until all image frames and sample IDs have been processed. The corrected rectangular sub-images are automatically renamed according to their sample IDs and the original image filenames (or timestamps), and saved to a separate subfolder named after the sample ID (e.g., all time-series images of "ID1" samples are stored in the "ID1" folder). This step segments and reconstructs the mixed high-throughput time-series image dataset into a clearly organized, normalized, corrected sub-image sequence based on sample IDs.

[0042] S24. Analysis of Region Coordinate Transformation and Saving. While performing perspective correction on each sample in S22, extract the "global coordinates of the analysis region" determined in S14 from the calibration instruction set in step S15. Using the forward perspective transformation matrix M calculated for the current sample in S22(2), map the global coordinate point set of the analysis region from the original image coordinate system to the corrected standard rectangular plane coordinate system through the perspective transformation function to obtain the "corrected analysis region coordinates". Save the "corrected analysis region coordinates" together with the sample ID as a new independent calibration file (e.g., save it as "1.json" in the "ID1" folder). This file will serve as the sole basis for precise cropping in step three.

[0043] Step 3: Incremental extraction of root growth trajectory. From the statically corrected sub-image sequence generated in S23, the temporal difference method is used to process the corrected sub-image sequence to obtain a difference image sequence. The difference image sequence is then accumulated pixel by pixel to obtain a static accumulated image within the observation period, from which trajectory information reflecting the actual root growth process is extracted. The specific implementation is as follows:

[0044] S31. Load the Region of Interest (ROI) and crop the image. For each independent image sequence generated in step S23 (e.g., the "ID1" folder), read the calibration file (i.e., [ID].json) generated in S24 within that folder, which contains the "corrected ROI coordinates". In the subsequent processing loop of S32, after loading each frame of the corrected sub-image in the time series, the image is cropped immediately according to this coordinate file, retaining only the image data within the ROI. All subsequent calculations (S32 to S34) are performed only on this cropped small-sized image, eliminating interference from irrelevant regions and improving computational efficiency.

[0045] S32: Using the temporal difference method, for each independent image sequence cropped in S31, the pixel-level difference between two adjacent frames is calculated in chronological order to obtain the difference image sequence. Specifically: (1) Select an independent cropped image subsequence arranged in chronological order for processing. Starting from the second frame of this time sequence, select two adjacent frames as a pair for pixel-level difference calculation. For example, first calculate the difference between the second frame and the first frame, then calculate the difference between the third frame and the second frame, and so on.

[0046] (2) For each selected pair of adjacent images, perform pixel-level subtraction. To retain only the region where pixel values ​​increase (i.e., the newly grown region), set all pixel values ​​less than 0 in the calculation results to 0. This operation can effectively suppress brightness reduction artifacts caused by fluctuations in lighting conditions.

[0047] (3) Combine the calculation results of all pixel positions to form a new image, namely the "difference image". This image mainly shows the areas that changed during the time interval between the two frames of images, while the pixel values ​​of the static background areas that did not change are close to zero in this image.

[0048] (4) Traverse the sequence: Repeat the above steps until all adjacent frame pairs in the entire image subsequence have been calculated.

[0049] This method can effectively suppress static backgrounds and highlight only pixel areas that change due to root growth over time intervals.

[0050] S33: To filter out artifacts caused by camera sensor noise or slight fluctuations in ambient light, the system performs threshold filtering on the differential image sequence in S32, retaining only pixels whose brightness changes exceed a preset threshold as effective growth regions. An accumulation matrix of the same size as the cropped image is maintained, and the effective growth regions at each time step (the result of S32(2)) are superimposed onto this accumulation matrix.

[0051] S34: After processing each pair of adjacent images, the current accumulation matrix is ​​converted into a grayscale image, serving as the cumulative growth trajectory map at that moment. This operation generates an image sequence of cumulative growth trajectories for each petri dish, where each image in the sequence contains the total growth trajectory from the start of the experiment to the current moment. A static accumulation image is reconstructed that represents the complete growth trajectory of the roots throughout the entire observation period. Independent processing of all samples significantly improves the overall efficiency of high-throughput data processing.

[0052] Step 4: Taking the primary root growth trajectory as the research object, perform skeleton extraction and length quantization. Perform adaptive binarization segmentation on the static accumulated image, and extract the centerline skeleton with a single pixel width based on the image after adaptive binarization segmentation to obtain the skeleton image. Use the shortest path algorithm to traverse all skeleton segments in the skeleton image to obtain the quantized primary root length at each time point. Specifically, this includes the following steps: S41: The image sequence of the cumulative growth trajectory generated in step S34 (i.e., each .png image generated over time in the folder) is used as the analysis object. The system traverses each frame of this sequence and independently executes subsequent steps S41 to S44 for each frame to obtain the principal root length at each time point. For each trajectory image currently traversed, the Otsu's method is used for adaptive binarization segmentation, converting the grayscale image into a binary image, where foreground pixels represent the root system and background pixels represent non-root regions. The algorithm ultimately selects the grayscale level that maximizes the inter-class variance between the foreground and background classes as the globally optimal threshold, thereby achieving stable and accurate segmentation of images with different lighting and contrast without manual intervention.

[0053] S42: Morphological Enhancement. A morphological closing operation is performed on the binary image obtained in S41 to fill in the tiny pores inside the root system and connect any potential breakpoints. The specific implementation of this step is as follows:

[0054] (1) Define the structural element (Kernel): First, define a 3x3 pixel square structural element.

[0055] (2) Perform 3-fold dilation: Using the 3x3 structuring element, perform 3 consecutive dilation operations on the binary image. This step can significantly thicken the root region, allowing its broken parts (breakpoints) to connect with each other, and completely fill the internal holes (noise).

[0056] (3) Perform three erosions: Subsequently, perform three consecutive erosion operations on the dilated result image using the same 3x3 structuring element. The purpose of this step is to "slim down" the dilated and thickened roots back to near their original thickness, while preserving the hole filling and breakpoint connection effects completed during the dilation process.

[0057] The combination of "three dilations followed by three erosions" constitutes a powerful morphological closing operation, which results in an enhanced binary image with a more complete root system morphology and better connectivity, laying a solid foundation for the subsequent skeletonization extraction of S43.

[0058] S43: Perform a skeletonization algorithm on the morphologically enhanced binary image, refining it into a single-pixel-width centerline that preserves the original topological structure—the skeleton. The length of this skeleton can accurately represent the curve length of the root system. The skeletonization algorithm simplifies a two-dimensional planar graphic into its one-dimensional structural representation, refining it into a "skeleton" with a width of a single pixel while fully preserving the original object's topological structure (such as connectivity, branches, and loops). This provides the foundation for subsequent path analysis and length measurement.

[0059] The identification of principal roots and the quantification of path search include: (1) performing connected component analysis on the skeleton graph in S43, identifying and separating all unconnected skeleton segments.

[0060] (2) For each independent skeleton segment, first find the two endpoints with the greatest Euclidean distance among all its pixels, and use them as the starting point and ending point of the path search respectively.

[0061] (3) Treat the skeleton segment as a graph structure composed of pixels, and use a cost-based graph path search algorithm to calculate the optimal path between the two endpoints along the skeleton pixels. Specifically, the travel cost of the skeleton pixels can be set to a very low value (e.g., 1), and the travel cost of the non-skeleton pixels can be set to a high value. The path found by the algorithm is the path with the lowest cost between the two points, which is the skeleton path. The length of the path is the number of pixels contained in the path.

[0062] (4) Traverse all skeleton segments, take the number of pixels of the longest path as the length of the skeleton segment, and finally identify the skeleton segment with the longest path as the main root.

[0063] S44: Physical Length Conversion and Recording. Based on the scaling factor k established in step two, the pixel length Lp of the skeleton segment identified as the principal root is converted into a physical length L = Lp / k in millimeters. This physical length is then associated with and recorded against the currently processed image frame (i.e., time point). For example... Figure 3 As shown. Simultaneously, the root length of the object under study can be automatically extracted according to actual analytical needs.

[0064] Step 5: Growth Dynamics Analysis and Visualization. The quantified primary root length data obtained from all time points recorded in S45 are correlated with the corresponding collection timestamps to generate a growth curve of primary root length changing over time, thereby achieving a quantitative assessment and visualization of root growth dynamics.

[0065] The automated measurement method of the present invention includes the following steps. (Refer to...) Figure 2 The system implementing this method may include an image acquisition device and a data processing system (such as a personal computer, server, etc.), wherein the processing system has deployed software modules for executing the method of the present invention, including:

[0066] The image calibration module acquires a high-throughput time-series image dataset of plant roots; selects a calibration image and generates a calibration instruction set; the calibration instruction set includes a custom identifier for each petri dish in a single image, the coordinates of the four vertices of the reference object, the global coordinates of the analysis area, and the physical size information of the reference object.

[0067] The correction and segmentation module is used to process image data in a high-throughput time-series image dataset based on the parameters of the reference objects in the calibration instruction set, and obtain corrected sub-image sequences.

[0068] The trajectory module is used to process the corrected sub-image sequence using the temporal difference method to obtain the difference image sequence; the difference image sequence is accumulated pixel by pixel to obtain the static accumulated image within the observation period; adaptive binarization segmentation is performed on the static accumulated image, and the centerline skeleton with a single pixel width is extracted based on the image after adaptive binarization segmentation to obtain the skeleton image; the shortest path traversal algorithm is used to traverse all skeleton segments in the skeleton image to obtain the principal root length obtained by quantization at each time point.

[0069] The quantitative analysis module generates a growth curve of the main root length over time based on the quantified main root length at each time point, thereby enabling quantitative analysis of the root growth dynamics.

[0070] The modules in the aforementioned high-throughput plant root dynamic growth phenotype measurement system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0071] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to implement the steps in an embodiment of a high-throughput measurement method for dynamic plant root growth phenotypes. Specific implementation methods can be found in the method embodiments, and will not be repeated here.

[0072] Furthermore, the present invention also provides a non-transitory computer-readable storage medium containing instructions, on which a computer program is stored. For example, a memory containing instructions that can be executed by a processor of a computer device to perform the above-described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc. When the computer program is executed by the processor, it can implement the steps in an embodiment of a high-throughput measurement method for dynamic plant root growth phenotypes. Specific implementation methods can be found in the method embodiments, which will not be repeated here.

[0073] Those skilled in the art will understand that embodiments of the present invention can provide methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0074] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0075] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0076] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0077] It should be noted that the specific embodiments described above enable those skilled in the art to more fully understand the present invention, but do not limit the present invention in any way. Therefore, although the present invention has been described in detail in this specification and embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the present invention; and all technical solutions and improvements that do not depart from the spirit and scope of the present invention are covered within the protection scope of the present invention patent. No reference numerals in the claims should be construed as limiting the scope of the claims. Any simple variations or equivalent substitutions of technical solutions that can be readily obtained by those skilled in the art within the scope of the technology disclosed in the present invention are within the protection scope of the present invention.

Claims

1. A method for measuring the dynamic growth phenotype of plant roots in high-throughput measurements, characterized in that, Includes the following steps: Acquire a high-throughput time-series image dataset of plant roots; select calibration images and generate a calibration instruction set; the calibration instruction set includes a custom identifier for each petri dish in a single image, the coordinates of the four vertices of the reference object, the global coordinates of the analysis region, and the physical size information of the reference object; Based on the parameters of the reference objects in the calibration instruction set, the image data in the high-throughput time-series image dataset is processed to obtain a corrected sub-image sequence. Specifically, a double-loop structure is used to perform batch perspective correction and segmentation on the samples. The outer loop is started to read each frame of high-throughput time-series image containing multiple culture dishes, and the inner loop is started to read the parameters of the corresponding calibration instruction set to obtain the IDs of different plant samples, the source coordinates of the reference vertex, and the global coordinates of the analysis area. For each sample ID, a forward perspective transformation matrix is ​​calculated based on the source coordinates of the reference vertex and the coordinates of the preset target rectangle. The forward perspective transformation matrix is ​​then used to process the high-throughput time-series image, cropping out the irregular areas of the sample and correcting them into standard rectangular sub-images. Using the perspective transformation matrix, the global coordinates of the analysis region of the sample are mapped from the original image coordinate system to the standard coordinate system coordinates of the corrected sub-image, thus obtaining the corrected sub-image sequence; The corrected sub-image sequence is processed using the temporal difference method to obtain the difference image sequence; The differential image sequence is accumulated pixel by pixel to obtain a static accumulated image within the observation period; The static accumulated image is subjected to adaptive binarization segmentation, and a centerline skeleton with a single pixel width is extracted based on the image after adaptive binarization segmentation to obtain a skeleton image. The shortest path algorithm is used to traverse all skeleton segments in the skeleton image to obtain the principal root length obtained by quantization at each time point. Based on the quantified taproot length at each time point, a growth curve of taproot length over time is generated, enabling quantitative analysis of root growth dynamics.

2. The method for measuring the dynamic growth phenotype of plant roots in high throughput according to claim 1, characterized in that, The process of selecting a calibration image and generating a calibration instruction set includes the following steps: Obtain a representative calibration image containing one or more culture containers; through the interactive interface, mark a reference object with at least four identifiable coplanar vertices on the image for each culture container, input the actual physical size of the reference object, and record the pixel coordinates of the four vertices of the reference object in the original image; Based on the pixel coordinates of the four vertices of the reference object and the coordinates of a preset standard target rectangle, a transformation matrix for correcting perspective distortion is obtained. Based on the transformation matrix, an analysis region for root growth is accurately selected. A matrix for inverse transformation is calculated, and the selected analysis region for root growth is converted back to its corresponding position in the original image coordinate system using the inverse transformation matrix. Save the user-defined identifier for each petri dish, the coordinates of the four vertices of the reference object, the coordinates of the analysis area, and the physical size information of the reference object as a calibration instruction set.

3. The method for measuring the dynamic growth phenotype of plant roots in high-throughput according to claim 1, characterized in that, The process of using the shortest path algorithm to traverse all skeleton segments in the skeleton image and obtain the quantized principal root length at each time point specifically includes the following steps: Connectivity analysis was performed on the skeleton to identify all independent skeleton segments; Calculate the farthest Euclidean distance between any two points in the skeleton image, and use the two points obtained as candidate start and end points of the path; The shortest path algorithm is used to calculate the shortest distance between the starting point and the ending point under the constraints of the skeleton segment; all skeleton segments are traversed, and the longest path calculated is identified as the main root path of the plant. The number of pixels contained in the main root path is counted, and the length of the main root is calculated.

4. The method for measuring the dynamic growth phenotype of plant roots in high throughput according to claim 1, characterized in that, The static accumulated image is adaptively binarized and segmented using the Otsu's method.

5. The method for measuring the dynamic growth phenotype of plant roots in high throughput according to claim 1, characterized in that, It also includes, before generating the growth curve of the principal root length changing over time, calculating the scale factor based on the parameters of the reference object of the calibration instruction set and the correction target pixel size, converting the pixel length of the skeleton segment identified as the principal root into the principal root physical length, and generating the growth curve of the principal root length changing over time based on the principal root physical length.

6. The method for measuring high-throughput plant root dynamic growth phenotypes according to claim 1, characterized in that, The method also includes performing a morphological closing operation on the binary image before extracting the centerline skeleton with a single pixel width based on the binary image to obtain an enhanced binary image, and extracting the centerline skeleton with a single pixel width from the enhanced binary image; wherein, performing the morphological closing operation on the binary image specifically involves performing three consecutive dilation operations on the binary image, and then performing three consecutive erosion operations on the dilated binary image to obtain the enhanced binary image.

7. A high-throughput measurement system for dynamic plant root growth phenotypes, characterized in that, include: The image calibration module acquires a high-throughput time-series image dataset of plant roots; selects a calibration image and generates a calibration instruction set; the calibration instruction set includes a custom identifier for each petri dish in a single image, the coordinates of the four vertices of the reference object, the global coordinates of the analysis area, and the physical size information of the reference object; The correction and segmentation module is used to process image data in the high-throughput time-series image dataset based on the parameters of the reference objects in the calibration instruction set to obtain a corrected sub-image sequence. Specifically, it performs batch perspective correction and segmentation on samples through a double loop structure. The outer loop reads each frame of high-throughput time-series image containing multiple culture dishes, and the inner loop reads the parameters of the corresponding calibration instruction set to obtain different plant sample IDs, reference vertex source coordinates, and global coordinates of the analysis area. For each sample ID, a forward perspective transformation matrix is ​​calculated based on the reference vertex source coordinates and the coordinates of a preset target rectangle. The forward perspective transformation matrix is ​​then used to process the high-throughput time-series image, cropping out irregular areas of the sample and correcting them into standard rectangular sub-images. Using the perspective transformation matrix, the global coordinates of the analysis region of the sample are mapped from the original image coordinate system to the standard coordinate system coordinates of the corrected sub-image, thus obtaining the corrected sub-image sequence; The trajectory extraction module is used to process the corrected sub-image sequence using the temporal difference method to obtain a difference image sequence; The differential image sequence is accumulated pixel by pixel to obtain a static accumulated image within the observation period; The static accumulated image is subjected to adaptive binarization segmentation, and a centerline skeleton with a single pixel width is extracted based on the image after adaptive binarization segmentation to obtain a skeleton image. The shortest path algorithm is used to traverse all skeleton segments in the skeleton image to obtain the principal root length obtained by quantization at each time point. The length quantification module generates a growth curve of the main root length over time based on the quantified main root length at each time point, thereby realizing a quantitative analysis of the root growth dynamics.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is loaded by the processor, it is able to perform the steps of the method according to any one of claims 1 to 6.

Images